Recall versus Recognition for Search
My last post talked about recall and recognition from the angle of letting a user find their friends. It would be naive however to presume that one is always better than the other. There are some notable areas where recall is clearly better.
Yahoo 1999
The Yahoo! homepage has always been optimised for recognition. This design evolved from a time when the web was a barren place. No one knew what was available, so the Yahoo! team decided to simply show what was out there. If you saw a term like “Soccer”, “Literature”. or “Lakers release Rodman”, you’d click it, but it would never occur to you to search for those, as you didn’t know they were there. Sure Yahoo! had a search box, but that wasn’t what they were about in the early. The core concept of Yahoo! was in categorizing and subcategorizing the web. It’s hard to let go of such a thing.
Categories and subcategories were the right design™ for a time when the users didn’t know what was available. The problem was that once we accepted that the web had pretty much everything, Yahoo! was left with shitload of categories no one cared about and a weak unloved search algorithm.
Google 1999
Google offer only recall. You have to know what you’re looking for. Google makes no attempt to let you “browse” the web. Recall is faster than recognition in these cases, but only works when users aren’t blindly searching to see what’s there. In 1995 this design would not have worked for the Google. The experience would have been poor. “I wonder does the web have anything about League of Ireland football? No. I wonder does it have any information on the when Pearl Jam are playing in Dublin. No. I wonder does it have any information on Eric Cantona. Yes, but very little”. Recall needs an exhaustive data set before it works well.
Searching versus Browsing
This is why I often advise e-commerce stores not to use search engine unless they’re willing to invest significantly. A weak search engine can be added easily using Sphinx or equivalent, however you have to be careful doing so. A bad search engine is worse than no search engine. I’ve seen the damage they can do. By providing a search engine, you’re telling your would-be customers to use it. Here are three problems I’ve seen break when a search engine is blindly added to a store.
Users don’t know what you stock
For the same reasons that Google would have struggled in 1996, your search engine can cause problems. You might have a wide selection of kitchen products, but if a user searches for “Tefal slow cooker” you may show zero results, even though you’ve a good selection of slow cookers and equivalents in your kitchenware section. After a couple of zero result searches you’ll inevitably lose your customer, when they would have been happy to browse and purchase from your kitchen category. A study by UIE showed that after a failed search 47% of users left the site. After a second failed search, another 30% followed them. It’s a risky business. Obviously smarter search engines can do a better job by showing alternates, but that’s where the significant investment comes in.
Analytics consultants will be falling over themselves to tell me “but you can check your search logs and see what you should stock“, but like A/B testing and other top tips, you need a significant amount of traffic before you can make such decisions.
Many products have bad names
Often this is beyond your control. Let’s have a look at some household brands with names not even a mother could love.
Sony Ericsson
Samsung
Nokia
All of these provide search engines enhanced with a “jump to” feature. Unfortunately the expected search terms are quite foreign to most users. Most Nokia users have no clue what model their phone is, I have no idea what Dell monitor I’m typing this on. A search engine doesn’t help here (unless it’s really smart, which doesn’t come for free)
Search is used for more than products.
By providing a search engine you send a message saying “type your query in here”. That won’t always work as you might be searching only through products, advertising new phones to frustrated users.
In closing
I’m not saying that Search is wrong, far from it. It’s by far the quickest way to get your customers where they want to go, provided they know where they’re going, and you can take them there. For e-commerce this isn’t always the case, I’ve worked on some shopping sites where it’s less than 10% of the cases. The mis-conception with search is that it’s easy and cheap/free to provide. As is the case with all white-label solutions, you can get a bare bones useless un-styled solution free. The rest will cost you.
If you are to provide a search engine, here are three areas I’ve learned to spend time on to avoid the common pitfalls.
- If your search box covers both products and customer support, you should differentiate them in search results and let the user switch result type.
- There is a difference between a search and a “jump to”. A “jump to” can uses tricks like auto-complete to bypass results pages and take a user directly to the product/page. Your search box should support this, but not demand it.
- Your search results page is a marketing page. You should spend time making ensuring that it presents products well, and provides key information to inform a customers (e.g. price, delivery date, availability etc). Displaying results in the well known Google style works, but is missing a good opportunity to present your products as well as possible. Even Google are starting to move away from the Google style displayof results. If you were to design a search engine for recipes, or supermodels, or baseball players would you really use text only? So why do it for products?
- Spend time designing your empty results page, it will be seen far more often than you’d like, no matter how hard you work. This is a good place to consider techniques such as alternate search recommendations.
Some worthwhile reading, if you find yourself working in this space. Thanks for reading.
- Producing great search results part one – Jared Spool introduces the issues with run of the mill search engines.
- Producing great search results part two – A discussion of what information needs to be provided on the results page.
- Design treasures from the Amazon – A peak into why Amazon.com works so well, and more importantly why you’re not Amazon.com
Follow @destraynor on Twitter.